---
name: "Wisdom of the Crowd" (football-data.co.uk) tested — the edge is LINE SHOPPING, not any single book, and B365 alone has none
description: 2026-07-25. Bet-when-a-book-beats-Pinnacle-fair-by-2%, tested on ~48k fixtures. B365 alone shows NO edge at any margin (0% to 8%, all t≈0). Best-of-all-books (MaxC) gives +1.39% t=+1.21 — not significant, and needs ~40 accounts. A first run showed +6-7% t=+5-8 but that was LOOK-AHEAD (book OPENING vs Pinnacle CLOSING).
metadata:
  type: project
---

User asked why football-data.co.uk's "Wisdom of the (Pinnacle) Crowd" appears to
work when xlow does not. Reconstructed the method — devig Pinnacle to get "true"
odds, bet another book offering >= 2% above that — and tested it on ~48,000
fixtures of `league_data/`.

## THE LOOK-AHEAD TRAP (I fell into it first)

Initial run: B365 **+6.18%** (t=+5.12), best-available **+7.07%** (t=+8.62).
Spectacular, and wrong. `B365H` is the OPENING price; `PSCH` is Pinnacle's
CLOSING price. Comparing them asks "did the opening price beat the line's eventual
resting place" — information unavailable at bet time. **Always match opening-to-
opening or closing-to-closing.** The `C` infix means closing throughout
football-data's schema.

## LIKE-FOR-LIKE RESULTS

**Bet365 closing vs Pinnacle closing — no edge at any threshold:**

| margin | bets | ROI | t |
|---|---|---|---|
| 0% | 8,183 | −0.76% | −0.37 |
| 2% | 3,706 | −2.11% | −0.63 |
| 5% | 1,331 | +0.67% | +0.10 |
| 8% | 576 | −4.02% | −0.38 |

No trend across thresholds — the signature of nothing.

**Best-of-all-books (MaxC):** 2% margin → n=30,191, **+1.39%, t=+1.21**, 95% CI
[−0.9%, +3.7%]. 5% margin → +2.47%, t=+1.18. Positive but NOT significant even at
n=30k, and it requires taking the highest price across ~40 bookmakers every time.

**So the value lives in LINE SHOPPING, not in any book being soft.**

## Why it could work for him and not for us — the venue, not the rule

At a bookmaker the posted odds ARE the price; a +1.4% edge is +1.4% kept. On an
exchange you cross a bid-ask spread to enter — 0.67c on a 25c contract, ~2.7% —
so the same +1.4% becomes −1.3%. Same signal, opposite sign, purely by venue.
This is the single cleanest statement of why everything in
[[project_xlow_spread_verdict]] died.

## Caveats that matter

- The page makes **no ROI claim** and he **stopped publishing selections in
  2025/26** when Pinnacle discontinued the free odds API. "It works for him" is
  unverified.
- Odds in `league_data` are UK/international Bet365. **US Bet365 prices worse**,
  so a US bettor starts below this baseline.
- Soft books limit/close winning accounts within months — the known ceiling on
  value betting, and the reason it does not scale.
- The user is US-based (Kalshi + Poly-US), so broad soft-bookmaker access — the
  thing that makes the strategy viable — is largely unavailable. That constraint
  is what pushed them onto exchanges originally.

Related: [[project_xlow_spread_verdict]], [[project_edge_mine_exhausted]],
[[project_nonprice_features_dead]].